User Theories about the Spotify Music Recommender System and Enhanced Explainable AI (XAI)

This research study is being conducted by Michael Ridley, a Ph.D. Candidate at Western University under the supervision of Dr. Jacquelyn Burkell, Faculty of Information and Media Studies.

The purpose of the study is to determine how explainable AI (XAI) can be enhanced for the lay population by exploring the folk theories held by users of the Spotify music streaming and recommendation system. Folk theories (aka mental models) of how Spotify makes its personalized recommendations may not be accurate but they typically are functional for the user. As a result these folk theories tell us where user beliefs might influence (negatively or positively) system understanding, user trust and perceptions of system accountability.

We are recruiting 10 developers of recommendation systems or developers involved in XAI initiatives. These participants are asked to review these folk theories and to identify XAI strategies or techniques that would enhance explainability to the lay population.

Prior to the interview, participants will be given a document describing the folk theories of Spotify users obtained during a series of interviews. Participants will be asked to read and reflect on this document before the interview. The Zoom interview should take approximately 60 minutes. Participants will be asked to explore both general and specific aspects of the folk theories and to suggest ways that XAI strategies or techniques could be utilized to both reinforce accurate belief, mitigate inaccuracies, and address other perceptions that could influence understanding, trust, and system accountability.

Participants will be compensated $25.

The interview data and other study data will not have any information identifying participants. The names of participants will not be used anywhere in the study or documents resulting from the study.  All personal identifiers will be removed from study data. Only anonymized data will be shared. 

Participation in this study is voluntary. Even if a participant consents to participate they have the right to not answer individual questions or to withdraw from the study at any time.

Michael Ridley
PhD Candidate
Faculty of Information and Media Studies
Western University
London, Ontario Canada
mridley2@uwo.ca or mridley@uoguelph.ca